Search

CN-121984093-A - Photovoltaic energy storage system switching control method based on island effect pre-judgment

CN121984093ACN 121984093 ACN121984093 ACN 121984093ACN-121984093-A

Abstract

The invention discloses a photovoltaic energy storage system switching control method based on island effect pre-judgment, which comprises the steps of firstly collecting various parameters of a power grid in real time, establishing a power grid state database, adopting a sensor to continuously monitor the various electrical parameters of the power grid, secondly, utilizing a machine learning algorithm to take data in the power grid state database as training samples, constructing an island effect pre-judgment model, and setting a pre-judgment probability threshold value, and sending switching signals of grid connection and grid disconnection when the pre-judgment probability of the island effect pre-judgment model exceeds the threshold value, and thirdly, calculating the required energy storage output adjustment according to the current load demand and the output power of a photovoltaic cell after receiving the switching signals. The charging and discharging states of the energy storage system are adjusted by controlling the energy storage converter, so that the sum of photovoltaic power and energy storage output force can be matched with load demands.

Inventors

  • CHEN HONGTIAN
  • TANG DING
  • YUAN ZHANPENG
  • HUANG YAOWEI
  • GUO QIYU
  • LIU JIAZHAO

Assignees

  • 广东数能优费电力科技有限公司

Dates

Publication Date
20260505
Application Date
20260204

Claims (7)

  1. 1. A photovoltaic energy storage system switching control method based on island effect pre-judging is characterized by comprising the steps of firstly collecting voltage, frequency and current parameters of a power grid in real time, establishing a power grid state database, adopting a sensor to continuously monitor all electric parameters of the power grid, secondly utilizing a machine learning algorithm to take data in the power grid state database as training samples to construct an island effect pre-judging model, wherein the island effect pre-judging model is used for predicting power grid parameters monitored in real time and predicting the probability of power failure caused by power failure of the power grid in a future period, thirdly, setting a pre-judging probability threshold value, sending a switching signal to determine to switch between grid connection and grid disconnection when the pre-judging probability of the island effect pre-judging model exceeds the threshold value, and fourthly, calculating the required energy storage output adjustment quantity according to the current load demand and the output power of a photovoltaic cell after receiving the switching signal, and adjusting the charge and discharge state of an energy storage system to enable the sum of the photovoltaic power and the energy storage capacity to be matched with the load demand.
  2. 2. The photovoltaic energy storage system switching control method based on island effect prediction according to claim 1, wherein in the first step, the sensor includes a voltage sensor, a current sensor, and a frequency sensor, and the monitored frequency is not lower than 50Hz.
  3. 3. The photovoltaic energy storage system switching control method based on island effect pre-judgment according to claim 1 is characterized in that in the second step, the machine learning algorithm comprises the following steps of firstly, carrying out feature extraction and pre-processing on collected power grid data, converting the power grid data into feature vectors suitable for being processed by a support vector machine, secondly, selecting a suitable kernel function, mapping a low-dimensional feature space into a high-dimensional feature space for nonlinear classification, and thirdly, training a support vector machine model through training samples, and determining parameters of an optimal hyperplane.
  4. 4. The photovoltaic energy storage system switching control method based on island effect pre-judgment according to claim 1, wherein in the second step, the machine learning algorithm is composed of an input layer, a hidden layer and an output layer, when the island effect pre-judgment is performed, each parameter of a power grid is used as input of the input layer, and the probability or classification result of the island effect is output by the output layer through multiple operation processing of the hidden layer.
  5. 5. The photovoltaic energy storage system switching control method based on island effect pre-judgment according to claim 1 is characterized in that in the second step, the machine learning algorithm comprises the following steps of randomly extracting a plurality of sub-sample sets from an original training sample set, wherein each sub-sample set is used for training a decision tree, in the second step, in the construction process of each decision tree, a part of features are randomly selected from all features as candidate features, and in the third step, optimal features are selected for splitting. Step four, inputting new power grid state data into each decision tree in the random forest to obtain a plurality of prediction results, and obtaining a final pre-judgment conclusion by integrating the results.
  6. 6. The photovoltaic energy storage system switching control method based on island effect pre-judgment according to claim 1, further comprising the step of monitoring voltage and frequency changes of the system in real time in the switching process, and further fine-tuning energy storage output and photovoltaic power through a feedback control mechanism to ensure stable operation of the switching process.
  7. 7. The photovoltaic energy storage system switching control method based on island effect pre-judgment according to claim 1, wherein in the fifth step, when the system is switched between a grid-connected state and an off-grid state, a disturbance signal is injected into the island effect pre-judgment model, when abnormal fluctuation of system voltage under the action of the disturbance signal is detected, when the frequency deviation rating exceeds 0.5Hz, the island effect risk is judged to exist, an instruction for compensating 30kW of power is sent to the energy storage system, and the energy storage battery pack is discharged at 30kW of power to perform power compensation.

Description

Photovoltaic energy storage system switching control method based on island effect pre-judgment Technical Field The invention relates to the field of photovoltaic energy storage, in particular to a photovoltaic energy storage system switching control method based on island effect prediction. Background With the rapid development of new energy industry, photovoltaic energy storage systems are widely used. The photovoltaic energy storage system can be operated in a grid-connected mode or off-grid mode, and the island effect is a problem to be solved in the switching process between the two operation modes. Island effect means that when the power grid stops supplying power due to faults or other reasons, the photovoltaic energy storage system cannot timely detect the state and still supplies power to surrounding loads to form an island. The safety of power grid maintenance personnel is threatened, power equipment can be damaged, and stable operation of the system is affected. At present, the existing grid-connected and off-grid switching control method of the photovoltaic energy storage system has the defect of coping with the island effect. Most of the methods are to detect and process after the island effect occurs, so that the island effect is difficult to be inhibited fundamentally, and the problems of voltage and frequency fluctuation and the like easily occur in the switching process, so that the normal operation of the load is influenced. Disclosure of Invention In order to solve the problems, the invention provides a photovoltaic energy storage system switching control method based on island effect pre-judgment, which comprises the steps of firstly collecting voltage, frequency and current parameters of a power grid in real time, establishing a power grid state database, adopting a sensor to continuously monitor all electric parameters of the power grid, secondly, utilizing a machine learning algorithm to take data in the power grid state database as training samples, constructing an island effect pre-judgment model, wherein the island effect pre-judgment model is used for predicting power grid parameters monitored in real time, predicting the probability of power failure caused by power failure of the power grid in a future period, thirdly, setting a pre-judgment probability threshold, sending a switching signal when the island effect pre-judgment model prediction probability exceeds the threshold, determining to switch between grid connection and grid disconnection, and fourthly, calculating the required energy storage output adjustment according to the current load demand and the output power of a photovoltaic cell after receiving the switching signal. The charging and discharging states of the energy storage system are adjusted by controlling the energy storage converter, so that the sum of photovoltaic power and energy storage output force can be matched with load demands. Further, in the first step, the sensor includes a voltage sensor, a current sensor, and a frequency sensor, and the monitored frequency is not lower than 50Hz. In the second step, the machine learning algorithm includes the steps of firstly extracting and preprocessing features of collected power grid data, converting the power grid data into feature vectors suitable for processing by a support vector machine, secondly selecting a proper kernel function, mapping a low-dimensional feature space to a high-dimensional feature space for nonlinear classification, and thirdly training a support vector machine model through training samples to determine parameters of an optimal hyperplane. In the second step, the machine learning algorithm is composed of an input layer, a hidden layer and an output layer, when the island effect is predicted, each parameter of the power grid is used as the input of the input layer, and the probability or classification result of the island effect is output by the output layer after multiple operation processing of the hidden layer. Further, in the second step, the machine learning algorithm comprises the following steps of firstly randomly extracting a plurality of sub-sample sets from an original training sample set in a put-back mode, wherein each sub-sample set is used for training a decision tree, secondly, randomly selecting a part of features from all the features as candidate features for each node in the construction process of each decision tree, and thirdly, selecting optimal features for splitting. Step four, inputting new power grid state data into each decision tree in the random forest to obtain a plurality of prediction results, and obtaining a final pre-judgment conclusion by integrating the results. Further, the method also comprises a step five of monitoring the voltage and frequency change of the system in real time in the switching process, and further fine-tuning the energy storage output and the photovoltaic power through a feedback control mechanism to ensure the stable operation